Martijn Pieters wrote :
What do you need load balancing for anyway? Is it for scalability, or for reliability?
I'm glad you asked, I think meny potential Zope users have their own answers, each with situation specifics. Mantra for IT components, "One (of anything) is not enough". For me, reliability is the most critical (as in MISSION CRITICAL) point. Scalability will never become an issue if reliability isn't there first. Why? Because Zope will never be adopted as a production system, without redundancy, failover, etc.
If you want scalability, surely the cost for ZEO can be justified?
Not really. There are two price points. A lot, and A lot * 2. I have expressed the need to start ZEO at a two server license, with a corresponding price. Then Zope could get the chance to prove it's worth, grow in percieved value, and to scale as needed. As things are, Zope hopefuls are sitting out because of missing redundancy, failover. That's also why I keep pointing out every possible means of simulating ZEO features on this list, at every opportunity. If one site get's Zope into production using some of these half measures, that's one more potential future ZEO customer. Otherwise, we'll never get far enough to "need" ZEO, in it's current form.
Also, do realize that Zope on a single machine can scale pretty awesome. We recon ZEO is only necessary for the absolute top range of websites.
That's another thing. Zope does great at throughput, but responsiveness isn't always up to expectations, for LAN based IntraNet users. A noticable speed improvment by load balancing across two ZEO servers could go a long way toward making a "full" ZEO license look "justified". Thanks for asking! Jerry S.